5 - Logistic regression results

Summary statistics of the dataset used in this analysis are presented in Table 1. Overall, 23 percent of the 3,248 staff in this sample were promoted between the 2003 and 2006 PBRF staff census. The data in Table 1 shows that the likelihood of being promoted was higher for staff who received higher grades in the 2003 Quality Evaluation. Fifty-five percent of staff with an ‘A’ were promoted, compared with 33 percent of staff with a ‘B’, 22 percent with a ‘C’ and 16 percent with an ‘R’.

The data in Table 1 also shows that staff who submitted a higher number of research outputs in the 2003 Quality Evaluation had a higher likelihood of being promoted. Around 40 percent of staff who submitted 40 or more research outputs were promoted, compared with 17 percent for staff that submitted less than 10 outputs.

However, to get a clearer picture of the association between the quality and quantity of research and the likelihood of being promoted, each of these factors needs to be examined while holding all other factors constant. The application of logistic regression to the dataset achieves this.

The results of four logistic regressions are presented in Table 2. Firstly, the factors associated with the likelihood of being promoted for all staff in the sample are analysed. Then, the factors associated with the likelihood of being promoted for each of the three academic ranks in 2003 are analysed. This allows for a more detailed look at the impact of the quality and quantity of research on the chances of promotion. Note that the results discussed in this section assume that all other factors are held constant while the association between a specific explanatory variable and the likelihood of being promoted is discussed.

The results show that there is a statistically significant association between the 2003 quality category received by staff and their chances of promotion. Staff who received higher quality categories had a greater chance of promotion. In the analysis of all staff, the odds of a staff member who received an ‘A’ quality category being promoted were over 17 times those of a staff member who received a ‘C’ quality category. Also, the odds of a staff member who received a ‘B’ quality category being promoted were 4.6 times those of a staff member with a ‘C’ quality category. Finally, the odds of a staff member with an ‘R’ quality category being promoted were 58 percent less than those of a staff member with a ‘C’ quality category.

The separate analysis of each position showed that the likelihood of being promoted from receiving a higher quality category tends to increase with academic rank. For example, the odds of a lecturer with a ‘B’ quality category being promoted were 3.1 times that of a lecturer with a ‘C’ quality category. However, the odds of a senior lecturer being promoted who had received a ‘B’ quality category were 5.6 times those of a senior lecturer with a ‘C’ quality category. Finally, the odds of an associate professor with a ‘B’ quality category being promoted were 7.4 times those of an associate professor with a ‘C’ quality category.⁵

The results also indicated that there is an association between higher research output and the chances of being promoted. In the analysis of all staff, one additional submitted research output increased the odds of promotion by 3.6 percent. For lecturers, this figure was 4.8 percent and senior lecturers 4.3 percent. However, an increase in submitted research outputs did not have a statistically significant effect on the chances of an associate professor being promoted. Therefore, it would appear that the association between higher research productivity and the likelihood of promotion diminishes for those in higher academic ranks.

In other results, the quadratic functional form for age suggests that the chances of promotion initially increase with age, but then decrease once a particular threshold age is reached. Figure 1 produces the predicted probabilities for a selected reference group (see note under Figure 1) to illustrate the relationship between age and the chance of promotion, while holding all other factors constant.

As can be seen, the chances of a lecturer being promoted were highest for staff aged 46, while for senior lecturers and associate professors the chance of promotion was highest at age 52.

Figure 1: Predicted probability of promotion by academic rank and age

Note:

The reference group used to generate the predicted probabilities was: male, experienced, received a ‘B’ quality category in 2003, employed at the University of Auckland, produced the average number of research outputs in that academic rank and in the ‘Biological sciences’ panel.

In Table 1, the actual percentage of men and women promoted was very similar, 27 percent and 26 percent, respectively. However, in the logistic regression analysis of all staff, the odds of women being promoted were around 20 percent less than for men, controlling for other factors. But the logistic regression analysis of each individual academic rank showed no statistically significant difference in the odds of men and women being promoted. Therefore, the evidence is mixed that there is any difference in the likelihood of men and women being promoted.

The results in Table 1 also show that less experienced staff had a lower chance of being promoted. The odds of staff who were identified as being new and emerging in the 2006 Quality Evaluation being promoted were 47 percent lower than staff not identified as new and emerging.

Overall, there was a greater likelihood of staff being promoted at the University of Waikato, University of Canterbury and Auckland University of Technology, compared with the University of Auckland.

The results also show that academic staff in subject areas such as ‘Business and economics’, ‘Education’, ‘Health’, ‘Humanities and law’ and ‘Social sciences’ had a greater likelihood of being promoted, compared with the reference category of ‘Biological sciences’.

Finally, the results of the analysis of all staff shows that lecturers were the academic rank most likely to be promoted, followed by senior lecturers and then associate professors. The odds of a senior lecturer being promoted were around 40 percent less than a lecturer. Also, the odds of an associate professor being promoted were around 97 percent less than that of a lecturer.

Given that the Gibson et al. (2008) study examined university economists, it is useful to do a similar analysis for this particular group. Logistic regression was applied to a dataset comprised solely of academics in the ‘Economics’ narrow subject area. The analysis showed that there were two key factors associated with a higher chance of promotion – a higher level of research quality and the rank of staff prior to promotion. The odds of staff who received a ‘B’ in the 2003 Quality Evaluation being promoted were 23 times those of staff who received a ‘C’ quality category. Also, the odds of senior lecturers and associate professors being promoted were much lower than for lecturers. However, there was no statistically significant association between the number of submitted research outputs in the 2003 Quality Evaluation and the likelihood of being promoted.

Footnote

5. The fact that lecturers who received an ‘A’ quality category had a similar likelihood of being promoted as a lecturer who received a ‘C’ is likely to be due to the small number of staff who received an ‘A’ quality category resulting in a lack of statistical significance.